Search for most updated materials ↑

 
The results of meta-analysis is presented in the form of graphs (graphs are plotted with studies against relative risk estimates). In this tutorial, a small spreadsheet application is presented to construct graphs for the presentation of the results of the meta-analysis. Download the graph maker.

This tutorial presents a general introduction to the process of meta-analysis. The web is replete with software and tutorials on meta-analysis. Some of them are listed in the following page.

Comments to current version of this lecture by Carl Michal

I have looked at the book by Petitti, and looked into your spreadsheet in a little more detail.  I believe I have found three errors in the spreadsheet, I was hoping you could confirm these.

1) In the spreadsheet, cell W10 contains the value of Petitti's D (on page 116 in her book).  But Petitti says that D=0 if Q<S-1.  The spreadsheet doesn't contain the logic to set D=0.  Replacing W10 with:
=if(u10/v10<0,0,u10/v10)
corrects this.

2) in column T of the data rows which is calculating 1/wstar as again on Petitti's page 116, the value of D (W10) is referenced, but the reference to W10 is not an absolute reference (ie it is W10 not $W$10) so that in subsequent rows, the D value does not appear in the formulae correctly.  In T13 for example, it appears as W11 rather than W10.

Correcting these two errors appears to correct the summary odds ratios.

3) In the calculation of the confidence bounds, it appears that the sum of wi rather than the sum of wstar is used (first line of Petitti's page 117).  In particular, cells I10 and J10 refer to H10. In place of H10,
they should 1/sqrt(u14)

With these three corrections, I now get results consistent with the results from R, and that are consistent (with small rounding error) with the example in section 7.8.2 of Petitti's book.

If you have any comments or questions, please send a message to supercoursehelp@gmail.com